We develop and empirically test a parsimonious model of how specific knowledge and the value of knowledge sharing influence manufacturing plants' incentive design choices. Our results confirm the prediction that increases in the extent of agents' specific knowledge and the value of knowledge sharing are associated with greater (less) reliance on output (input) performance measures. Moreover, consistent with our model's prediction, we find that plants rely more on group-based (as opposed to individual-based) output performance measures when the value of knowledge sharing is higher, or the extent of agents' specific knowledge is lower. Finally, consistent with previous research, we find that as output performance measures become noisier, firms rely less on these measures in incentive contracts.

This paper examines the length of time over which CEO performance is evaluated (the "performance period") in CEO performance-based equity awards (PBEAs). Departing from the primary emphasis of agency theory on moral hazard problems, we develop a model in which short performance periods are instrumental in sorting CEO talents. The model predicts that short performance periods are preferred when CEOs have low expected productivity or valuable alternative employment opportunities, and when firms face high operating uncertainty or high dispersion of managerial productivity. We find empirical support for these predictions in a sample of S&P 1500 industrial firms granting PBEAs to CEOs. We also document that CEO turnover is higher for underperforming CEOs with shorter performance periods, validating the sorting role of performance periods.